FAST IMAGE-RESTORATION FOR REDUCING BLOCK ARTIFACTS BASED ON ADAPTIVECONSTRAINED OPTIMIZATION

In this paper we propose a fast adaptive image restoration filter usin g DCT-based edge-classification for reducing block artifacts in compre ssed images. In order to efficiently reduce block artifacts, edge dire ction of each block is classified by using the DCT coefficients, and a n adaptive constrained least squares (CLS) filter along the edge direc tion is used for filtering the corresponding block. The proposed resea rch is based on the observation that degradation due to block artifact s is a nonlinear and many-to-one mapping operator. Then, we propose an approximated version of a constrained optimization technique as a res toration process for removing the nonlinear and space-varying degradat ion operator. For real-time application, the proposed restoration filt er is implemented in the form of a truncated FIR litter, which is suit able for postprocessing the images in real-time video systems such as HDTV or video conferencing systems. (C) 1998 Academic Press